背景
发现一个奇怪的内核 哎呀，谷歌了很多，一无所获。
背景:
内核版本为3.0.8
有两个进程，比方说p1，p2
p2有很多线程(大约30个)
p1 连续调用 system(pidof("name of p1"))
内核运行几天后可能会Oops。我发现的主要原因是 unwind_frame从 get_wchan 得到一个奇怪的框架->fp(0xFFFFFFFF)
执行这一行时
frame->fp = *(unsigned long *)(fp - 12);
CPU 将尝试访问 0xFFFFFFF3，并导致分页错误。
我的问题是:
上下文切换前fp寄存器是怎么保存的变成0xFFFFFFFF的？
这里是CPU信息
# cat /proc/cpuinfo
Processor : ARMv7 Processor rev 0 (v7l)
processor : 0
BogoMIPS : 1849.75
processor : 1
BogoMIPS : 1856.30
Features : swp half thumb fastmult vfp edsp vfpv3 vfpv3d16
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc09
CPU revision : 0
这是 Oops 和 pt 寄存器:
[734212.113136] Unable to handle kernel paging request at virtual address fffffff3
[734212.113154] pgd = 826f0000
[734212.113175] [fffffff3] *pgd=8cdfe821, *pte=00000000, *ppte=00000000
[734212.113199] Internal error: Oops: 17 [#1] SMP
--------------cut--------------
[734212.113464] CPU: 1 Tainted: P (3.0.8 #2)
[734212.113523] PC is at unwind_frame+0x48/0x68
[734212.113538] LR is at get_wchan+0x8c/0x298
[734212.113557] pc : [<8003d120>] lr : [<8003a660>] psr: a0000013
[734212.113561] sp : 845d1cc8 ip : 00000003 fp : 845d1cd4
[734212.113583] r10: 00000001 r9 : 00000000 r8 : 80493c34
[734212.113597] r7 : 00000000 r6 : 00000000 r5 : 83354960 r4 : 845d1cd8
[734212.113613] r3 : 845d1cd8 r2 : ffffffff r1 : 80490000 r0 : 8049003f
[734212.113632] Flags: NzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment user
[734212.113651] Control: 10c53c7d Table: 826f004a DAC: 00000015
调用栈如下:
[734212.117027] Backtrace:
[734212.117052] [<8003d0d8>] (unwind_frame+0x0/0x68) from [<8003a660>] (get_wchan+0x8c/0x298)
[734212.117079] [<8003a5d4>] (get_wchan+0x0/0x298) from [<8011f700>] (do_task_stat+0x548/0x5ec)
[734212.117099] r4:00000000
[734212.117118] [<8011f1b8>] (do_task_stat+0x0/0x5ec) from [<8011f7c0>] (proc_tgid_stat+0x1c/0x24)
[734212.117158] [<8011f7a4>] (proc_tgid_stat+0x0/0x24) from [<8011b7f0>] (proc_single_show+0x54/0x98)
[734212.117196] [<8011b79c>] (proc_single_show+0x0/0x98) from [<800e9024>] (seq_read+0x1b4/0x4e4)
[734212.117215] r8:845d1f08 r7:845d1f70 r6:00000001 r5:8ca89d20 r4:866ea540
[734212.117237] r3:00000000
[734212.117264] [<800e8e70>] (seq_read+0x0/0x4e4) from [<800c8c54>] (vfs_read+0xb4/0x19c)
[734212.117289] [<800c8ba0>] (vfs_read+0x0/0x19c) from [<800c8e18>] (sys_read+0x44/0x74)
[734212.117307] r8:00000000 r7:00000003 r6:000003ff r5:7ea00818 r4:8ca89d20
[734212.117340] [<800c8dd4>] (sys_read+0x0/0x74) from [<800393c0>] (ret_fast_syscall+0x0/0x30)
[734212.117358] r9:845d0000 r8:80039568 r6:7ea00c90 r5:0000000e r4:7ea00818
[734212.117388] Code: e3c10d7f e3c0103f e151000c 9afffff6 (e512100c)
[734212.113136] Unable to handle kernel paging request at virtual address fffffff3
[734212.113154] pgd = 826f0000
[734212.113175] [fffffff3] *pgd=8cdfe821, *pte=00000000, *ppte=00000000
[734212.113199] Internal error: Oops: 17 [#1] SMP
最佳答案
此错误已由 Konstantin Khlebnikov 修复，详细信息可在 git commit log 中找到.
关于c - unwind_frame 导致内核分页错误，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/18479894/